The tax deduction in tuition and fees (DTF) was implemented in 2001 as part of the Economic Growth and Tax Relief Reconciliation Act. Under this program, households could deduct tuition and fees paid for undergraduate or graduate education from gross income without needing to itemize deductions. Households are eligible to use the DTF based on adjusted gross income (AGI) net of all other above-the-line deductions. Beginning in 2004, the maximum deductions followed a tier system in which joint (single) filers with eligible income of less than $130,000 ($65,000) were eligible for a $4,000 maximum deduction, while households with eligible incomes of $130,000-$160,000 ($65,000-$80,000) were eligible for a $2,000 maximum deduction. Households above $160,000 ($80,000) were ineligible for the deduction. The change in tax liability is based on whether the household claimed DTF scaled by the marginal tax rate at their income level.
Hoxby and Bulman (2016) exploit income eligibility thresholds to estimate the impact of the above-the-line deduction of tuition and fees on enrollment. They use a regression discontinuity design but leave out observations right around the threshold to account for the possibility of manipulation in the running variable. This MVPF estimate considers the MVPF implied by the discontinuity faced by single filers with incomes near $65,000.
Hendren and Sprung-Keyser (2020) take the causal estimates from Hoxby and Bulman (2016) and project the impact of the tuition deduction on lifetime earnings and tax revenue. They utilize estimates from Zimmerman (2014) on the impact attendance of college on earnings and assume that the returns to college are constant in percentage terms over the lifecycle.
Pays for Itself
Hendren and Sprung-Keyser (2020) calculate net costs beginning with the direct cost of the tax deduction. This is estimated at $90.30, which is the average deduction claimed of $361 multiplied by the marginal tax rate of 25% at $65,000. They then include three additional fiscal costs. First, they estimate the cost of increased educational attainment, net of private tuition payments, using data from the Delta Cost Project in 2011. This yields an enrollment cost of $70.97. Second, they account for the changes in taxes paid and transfers received based on the earnings gain calculated in the willingness to pay section. This gives a $130.10 increase in government revenue. Finally, they incorporate changes in government educational spending as the tax deduction induces individuals to attend more (or less) costly institutions. These are costs are due to changes in the composition of colleges attended by those receiving the deduction. Hoxby and Bulman (2016) provide estimates of the change in the change in core educational costs and the change tuition paid by the students at the discontinuity. The difference between these two measures is the average change in government educational costs from compositional changes in schooling. Hendren and Sprung-Keyser (2020) calculate that this decreases costs by $481. Adding together these components yields a net government cost of $459.9
Hoxby and Bulman (2016)find that, at the discontinuity, college attendance increases by 0.053 percentage points amongst students that graduated high school in the previous year. To calculate the willingness to pay, Hendren and Sprung-Keyser (2020) transform the effect of college attendance into additional years of schooling by assuming that individuals who enroll complete two additional years of college education on average. They then use estimates from Zimmerman (2014) to estimate the impact on lifetime earnings, combined with an assumption of a constant earnings impact over the life cycle. They then calculate post-tax and post-transfer earnings by assuming a tax and transfer rate of 18.9% (based on estimates from the Congressional Budget Office). They then subtract out the individual contribution to tuition, which is estimated by the average tuition net of aid using data from the Delta Cost Project for 2006 (the midpoint between 2004-2008, the period studied). The projected earnings gains provide the willingness to pay for all individuals induced to change their behavior and receive more education as a result of the tax deduction. They calculate total willingness to pay by summing those earnings gains with the value of the transfer for the fraction of individuals not induced to change their behavior. The WTP is $485.23, driven by $491.0 in post-tax earnings and $95.6 in private tuition contributions and a transfer value for those who don’t change behavior.
The negative net cost and positive willingness to pay implies an infinite MVPF. But, the MVPF has a wide confidence interval of [-\infty,\infty]. For context, if the policy had no effect on college attendance or college costs, we would expect the MVPF of this policy to be around 1. (The policy would simply be a tax cut that induced a muted behavioral response.) As these calculations show, impact of the policy on schooling expenses is sufficiently large that the MVPF estimate deviates from meaningfully from 1. That said, confidence interval around the estimate is quite large and so we cannot rule out an infinite MVPF or an MVPF of 0. In other words, this means that one cannot statistically rule out that the policy either pays for itself or provides no benefit to the beneficiaries, and suggests the value of further work to increase the statistical precision of the impact of the policy.
Hendren, Nathaniel and Ben Sprung-Keyser (2020). “A Unified Welfare Analysis of Government Policies.” The Quarterly Journal of Economics, 135(3): 1209–1318. DOI: https://doi.org/10.1093/qje/qjaa006
Hoxby, Caroline M. and George B. Bulman (2016). “The Effects of the Tax Deduction for Postsecondary Tuition: Implications for Structuring Tax-Based Aid”, Economics of Education Review, 23-60. https://www.sciencedirect.com/science/article/abs/pii/S0272775715001326
Zimmerman, Seth D. (2014). “The Returns to College Admission for Academically Marginal Students.” Journal of Labor Economics, 32(4), 711-754. DOI: https://doi.org/10.1086/676661